摘要： The aim of this study was to develop a low-dose radiography protocol for the neonatal ICU (NICU) using a new mobile digital radiography system with advanced denoising image processing and to evaluate the noninferiority of that protocol. In this prospective randomized study, 40 neonates in the NICU underwent radiography of the thorax and abdomen with two different mobile radiography units: conventional technique on one unit (50 kV, 1.6 mAs, and no additional filtration) and a new technique on another unit (54 kV, 0.1-mm Cu filtration). Three low-dose protocols for the second unit were developed in a phantom study: protocol A (100% equivalent dose with conventional protocol), protocol B (80% equivalent dose), and protocol C (64% equivalent dose). The noninferiority of each low-dose protocol was assessed by three independent readers using image quality criteria. Forty patients each underwent three pairs of radiography examinations (protocol A and the conventional protocol, protocol B and the conventional protocol, and protocol C and the conventional protocol), except one pair that did not include one image of the conventional protocol. The interrater reliability among the three readers was 0.91 (p < 0.001). Both of the low-dose protocols (B and C) were statistically noninferior to the conventional protocol with respect to overall image quality. Protocol B better depicted almost all anatomic landmarks and had better overall image quality than the conventional protocol. Using appropriate technique and acquisition factors, radiation dose can be lowered on a digital radiography system without significant effect on the image quality by adding filtrations and a new denoising technique.